Lighting-Invariant Visual Odometry using Lidar Intensity Imagery and Pose Interpolation

Recent studies have demonstrated that images constructed from lidar reflectance information exhibit superior robustness to lighting changes in outdoor environments in comparison to traditional passive stereo camera imagery. Moreover, for visual navigation methods originally developed using stereo vision, such as visual odometry (VO) and visual teach and repeat (VT&R), scanning lidar can serve as a direct replacement for the passive sensor. This results in systems that retain the efficiency of the sparse, appearance-based techniques while overcoming the dependence on adequate/consistent lighting conditions required by traditional cameras. However, due to the scanning nature of the lidar and assumptions made in previous implementations, data acquired during continuous vehicle motion suffer from geometric motion distortion and can subsequently result in poor metric VO estimates, even over short distances (e.g., 5–10 m). This paper revisits the measurement timing assumption made in previous systems, and proposes a frame-to-frame VO estimation framework based on a novel pose interpolation scheme that explicitly accounts for the exact acquisition time of each feature measurement. In this paper, we present the promising preliminary results of our new method using data generated from a lidar simulator and experimental data collected from a planetary analogue environment with a real scanning laser rangefinder.